This thesis presents a scalable, generic methodology for microbial phenotype prediction based on supervised machine learning, several models for biological and ecological traits of high relevance, and the deployment in metagenomic datasets.